基于RED的网络拥塞控制算法研究及改进

Research and Improvement of Network Congestion Control Algorithm Based on Random Early Detection

随机早期检测(Random Early Detection,RED)算法是为TCP流设计的一个主动队列管理机制,能在一定程度上缓解网络拥塞.在RED算法中丢包率与包的大小之间是独立的,这就造成了对小包的歧视.在packet size算法中,通过对RED算法进行适当的改进,体现了丢包率和吞吐量之间的公平性.在此基础上进一步分析,用平均包的大小来影响丢包率以提高网络性能.仿真实验表明该算法对网络拥塞控制具有较好效果.

Random early detection （RED） algorithm is an active queue management mechanism for TCP flows, which can alleviate the network congestion in a certain extent. The packets dropping probability is independent from the packet size in RED algorithm, which will cause discrimination of the small packets. In packet size algorithm, through making suitable improvement to RED algorithm, we can get fairness between the packets dropping probability and the throughput. On the basis of further analysis, in order to enhance network performance, we use the average packet size to influence the packets dropping probability. Simulation experiments show that the improved algorithm has a better effect for the network congestion control.